International Journal of Language and Linguistics 2015; 3(5): 313-322

Published online September 16, 2015 (http://www.sciencepublishinggroup.com/j/ijll)

doi: 10.11648/j.ijll.20150305.16

ISSN: 2330-0205 (Print); ISSN: 2330-0221 (Online)

Internal and External Factors Affecting Learning English as a Foreign Language

Soheil Mahmoudi1, Asgar Mahmoudi

2, *

1Department of English Language, Ahar Branch, Islamic Azad University, Ahar, Iran 2Department of English Language, Ardabil Branch, Islamic Azad University, Ardabil, Iran

Email address: Soheilmahmoudi70@gmail.com (S. Mahmoudi), Mahmoudi301@gmail.com (A. Mahmoudi)

To cite this article: Soheil Mahmoudi, Asgar Mahmoudi. Internal and External Factors Affecting Learning English as a Foreign Language. International Journal

of Language and Linguistics. Vol. 3, No. 5, 2015, pp. 313-322. doi: 10.11648/j.ijll.20150305.16

Abstract: This study investigated the effects of internal and external factors on learning English as a foreign language from

Iranian EFL learners’ points of view. Copies of a 30-item Lickert-scale questionnaire, addressing internal and external factors

or principle components, were distributed among about 140 postgraduate students of ELT in three universities in Iran. The

collected data were then subjected to Principle Component Analysis (PCA). The findings revealed that while internal and

external components are distinguishable, many of the variables do not heavily load on the principle component to which they

theoretically belong. After separating the non-correlating variables it became clear that most of these variables are very

important variables. Further analysis indicated that it is possible to divide internal variables to cognitive and affective and

external variables to environmental and curricular. The conclusion reached was that the importance of variables should not be

judged based on their nature but based on the importance accorded to them by the respondents. It was also concluded that

extreme attention paid to internal variables should be balanced against external variables.

Keywords: Principle Component Analysis, Internal Variables, External Variables

1. Introduction

1.1. Overview

This survey study was designed to explore factors or

principle components affecting Iranian EFL learners’ success

from their own perspectives. The study also probed into these

components relationships with variables loading on them.

The roles of internal and external factors or components in

the acquisition of second or foreign language have been

broadly investigated in the past (e.g., Brown, 1995; Ellis,

2008; Nunan, 1988; Ortega, 2009). However, in many cases

the boundaries between these two principle components and

the way their respective variables load on them are left

unexplored. For example, while anxiety and attitude are

internal variables, they can be heightened or weakened by

external variables. On the other hand, shortcomings in the

external variables, such as unfavorable institutional context,

can be compensated for by the autonomy (an internal

variable) that a student has. Therefore, there might be

interactions between these two groups of variables and they

may reciprocally affect each other. Exploring these kinds of

relationships was the main concern of this research.

1.2. Statement of the Problem and Significance of the Study

Looking through research articles, journals, and books one

can see that countless research has been conducted on the

internal and external factors or principle components

affecting second or foreign language acquisition. However,

the question of the interactions between these two groups of

variables is given little attention. Robinson and Ellis (2008)

indicate that all these variables are inextricably intertwined in

a rich, complex, and dynamic way in languages. The purpose

of categorization, therefore, is only to understand the

situation better not to claim that these factors have nothing to

do with each other. Foreign language learners themselves too

have not been asked to express their views over these issues

very frequently, as the majority of research in the field has

been experimental in nature. This study aimed at bridging the

gaps in these two areas feeling that this might provide a

deeper understanding of what is going on in EFL

environments. Moreover, it tried to find out if it is correct to

look at the internal and external factors as separate groups of

variables having no interaction with each other. Still more,

the study tried to investigate the subdivisions of the internal

314 Soheil Mahmoudi and Asgar Mahmoudi: Internal and External Factors Affecting

Learning English as a Foreign Language

and external factors to see if they are further divisible into

cognitive and affective and curricular and environmental,

respectively.

1.3. Significance of the Study

A more profound understanding of the internal and

external principle components, and articulating them at some

detail with a reference to their due shares in developing

second or foreign language proficiency, will certainly help all

stakeholders to make wiser decisions. In particular

administrators and people in charge of directing the

educational system will know where to invest their time,

energy, and money. Students too can regulate their own

learning processes by the enhanced awareness that would

result. Many a time students do not know why they are not

making the progress they think they should with respect to all

their efforts. Parents, too, sometimes get frustrated by their

children’s sluggish progress because of the poor

understanding that they have of the so many variables

affecting their children’s performance. And even many

teachers get bogged down with all good intentions that they

bring to the process of teaching and learning. This study is

significant because it was designed to tell these people why

sometimes things go wrong. It is also significant because it

was an effort to provide clear understanding of the most

important variables that from learners’ perspectives affect

their learning of a foreign language, in this case English.

1.4. Design of the Study

This study was conducted following an ex post facto

design due to the fact that the variables and their clustering

and inter-correlations were thought to be present prior to the

beginning of the study. This means that, the study did not

involve any interference and merely tried to identify the

clustering and patterns of interactions that existed among the

foreign language learning variables from the respondents’

perspectives. Similar to experimental studies this design only

explains the consequent of a condition. Unlike experimental

studies, however, it does not let the researcher manipulate the

variables of the study.

2. Review of the Related Literature

2.1. Introduction

Looking at all variables affecting foreign language

learning simultaneously would be a little bit difficult. It

seems logical to classify these factors into two groups of

internal and external that include elements of the same type

(Madrid, 1995). External factors are largely dealt with in

books dealing with curriculum development and course

design (e.g., Dean Brown, 1995; Lewis & Hill, 1985; Nation

& Macalister, 2010; Nunan, 1988; Richards, 2001; White,

1988) while internal factors are discussed in SLA and

psychology books (e.g., Brown, 2007; Robinson & Ellis,

2008; Gass & Selinker, 2008; Ortega, 2009; Woolfolk, Winne,

& Perri, 2003).

2.2. Internal Variables

Internal variables imply cognitive and affective factors

such as motivation, intelligence, anxiety, risk-taking ability,

etc. Because of space limitation, only some of these variables

are elaborated on here.

Many studies have confirmed that motivation correlates

strongly with proficiency, indicating both that successful

learners are motivated and that success improves motivation.

Motivation has been recognized as an important variable

determining L2 achievement and attainment for a long time.

Motivation is believed to act as an engine generating learning

and then propelling students forward helping them overcome

the difficulties they encounter in learning a foreign language

(Cheng and Dörnyei, 2007; Dörnyei and Csizer, 1998).

Brown (2007) considers motivation as an affective factor that

plays a central role in learning a second or foreign language.

Cohen (2010) sees motivation as a dynamic process that is

not stable but is in a continuous change.

Language learners’ attitudes toward the language being

learned, likewise, can have a significant impact on SLA.

Where the community has a broadly negative view of the

target language and its speakers, or a negative view of its

relations to them, learning is typically much more difficult

(Gardner, 1985; “Attitude,” 2009). According to Siegel

(2003), motivation is affected by learners’ attitudes toward

the L2, its speakers, and the speakers’ culture.

Extraversion and introversion are two personality types

that fall within the brief of internal variables. Studies have

revealed that extraverts acquire a second language better than

introverts. Gregarious people usually tend to communicate

with others even if they are not sure they will succeed

(Kinginger and Farell, 2004).

2.3. External Variables

Among external variables one can refer to such variables

as social class, first language, teachers, early start, L2

curriculum, etc. Generally speaking, external variables can be

categorized into the two groups of environmental and

curricular but the list of external variables referred to here,

like the list of internal variables, is not exhaustive, as in

many other studies. Nation and Macalister (2010), for

example, in discussing the importance of curricular issues,

highlight the importance of needs analysis, sequencing the

course materials, evaluation, format and presentation of

materials. Richards (2001) too emphasizes the roles of

institutions, teachers, and learners in providing for effective

learning.

Teacher behavior definitely influences all kinds of learning

especially learning a foreign language. According to Cheng

and Dörnyei (2007), teachers can fire students’ enthusiasm by

being a personal model in the class. Stipek (2002), also

points to the importance of the teachers’ projection of

enthusiasm. With the development of technology the Internet

is playing a more and more important role in learning English.

English students are downloading English songs, and films

that let them get exposure to real English at a globalized

International Journal of Language and Linguistics 2015; 3(5): 313-322 315

communicational level (Nurul Islam, 2011). The Internet is

much more than this. All people around the globe, especially

students, use it to do research, to access library materials,

online quizzes, podcasts, and the like (Khanchali, & Ziadat,

2011).

It has been agreed on the fact that there should be some

sort of variety in EFL learning. One way of bringing variety

into EFL classes is the use of films (Ismaili, 2013). It is

revealed that L2 learners who do not possess the same

linguistic base as the L1 face a lot of difficulties. The

difficulties get bigger when there is a bigger difference

between L1 and L2 (Karim, 2003; Segalowitz 1986). One of

the ways to cope with this problem is integrating English

films, sometimes manipulated for pedagogical purposes, into

the teaching process.

3. Method

3.1. Introduction

In the case of EFL, a lot of questions can be asked that

roughly can be divided into questions addressing internal or

external variables. However, the loading of each of these

questions on the main construct and its subdivisions will not

be clear unless a PCA is run on the collected data. This is

because responses to questions or the importance assigned to

each variable in a questionnaire are situation and person

specific and vary from one situation and person to the other.

It is only after running a PCA that patterns of groupings in

that particular context emerge. On the other hand, researchers

might be interested in finding out if the principle components

they identify after running a PCA can be further subdivided

into additional specific components. This is possible in two

ways: impressionistically and by running additional PCAs.

Impressionistic grouping of variables has the danger that the

variables might not load on the component that the researcher

thinks they should even if they are conceptually and

theoretically related. Running a PCA, in contrast, increases

the precision and brings to the surface things that otherwise

might remain unnoticed.

3.2. Research Questions and Hypotheses

Following from what was said, this research tried to find

out if all of the variables in each category load on the

principle component to which they are attributed and if not

why. The study, in other words, tried to answer the questions

‘Do all variables that are theoretically considered to be

internal correlate significantly with each other? Do all

variables that are theoretically considered to be external

correlate significantly with each other? Are there any internal

and external variables that correlate significantly with each

other or with the other principle component? Are internal

variables divisible to cognitive and affective? Are external

variables divisible to curricular and environmental? How the

answers to these questions can be justified?

The null research hypotheses that were derived from these

questions are as follows:

H01: All variables that are theoretically considered to be

internal or external do not necessarily correlate significantly

with variables of their own type.

H02: There are no internal and external variables that

correlate significantly with each other or with the principle

component to which they do not belong.

H03: Internal and external variables are not treated, i.e.,

rated differently by EFL learners with respect to their

importance.

H04: Internal and external variables are divisible to

cognitive and affective and curricular and environmental,

respectively, with regard to the amount of importance that

students assign to them.

3.3. Participants

The participants of this study were all Master’s degree

students of English Language Teaching (ELT) in three

universities in northwest Iran, two in the provincial city of

Ardabil, and one in Ahar city. Naturally, all of the students

were above 22 years old and had an English language

learning experience of at least five years. A great majority of

the participants were fluent bilinguals of Persian and Azeri

but there were a few of them who did not know Azeri well.

No screening for proficiency was done before beginning of

the research because the research was not intended to

measure the participants’ gains in proficiency over time

rather to elicit their opinions about the importance of

variables affecting learning English as a foreign language.

3.4. Instruments

The instruments used in this study were of three types. The

first instrument was a 5-point Likert scale questionnaire

containing 30 questions. The questions were of two types

related to internal and external variables that could be further

sub-divided into cognitive and affective in the case of

internal variables and curricular and environmental in the

case of external variables. All of the questions were derived

from research articles and a rough balance was established

with respect to the number of questions addressing each

variable type. The values of responses to each question

ranged from 1 to 5. One represented the least effect and five

represented the most effect. The questionnaire was designed

to give the fullest possible coverage to the variables that,

according to the literature on the field, affect learning English

as a foreign language but the length of the questionnaire was

kept in control not to exceed the limit that might have

discouraged the respondents from answering all of the

questions with enough attention.

Another type of instrument used in this study was

Microsoft Office’s Excel spreadsheet that was used to

calculate the means of responses to each question in the

questionnaire. These means at the later stages of the study,

when the variables were divided into two components, were

used as distributions of mean scores to run an Independent-

samples T-test between the two groups of variables to

discover if according to students’ responses they were

316 Soheil Mahmoudi and Asgar Mahmoudi: Internal and External Factors Affecting

Learning English as a Foreign Language

significantly different from each other.

The last instrument used was the SPSS package that was

used to analyze the collected data. As a prerequisite of

descriptive studies, it was necessary to check for the

reliability of the questionnaire. Further, it was necessary to

run an exploratory PCA to find out if any variable, from

participants’ points of view, was exerting undue influence

upon foreign language learning. The findings would also be

much more understandable if they could be represented

diagrammatically. SPSS was used to do all these things.

3.5. Procedure

Since some respondents answer questions in a

questionnaire superficially, it was decided that if anyone’s

responses were the same for more than one-third of the

questions in the questionnaire, that copy of the questionnaire

be discarded from the study. For this reason, although

initially more than 160 copies of the questionnaire were

distributed among the MA students, the ultimate number of

copies used in the data analysis was 136.

Of the 30 questions in the questionnaire 15 were targeted

at the internal variables and 15 at the external variables but

not to let out the purpose of the study, the questions were

arranged in an odd-even order with odd questions addressing

the internal variables and even questions addressing the

external variables. The organization of the questions relating

to subdivisions of the internal and external variables was

random.

After the questions were answered and the data were

collected, they were entered into SPSS. SPSS recognizes

each question in a questionnaire as a single variable and deals

with it accordingly. After the data inputting and as the first

stage a Chronbach Alpha reliability test was run to find out if

the questionnaire was reliable. After that, a PCA with two

principle components on the overall data and then two more

PCAs on each major group of variables again with two

principle components were run with Scree and Component

plots to test the hypotheses of the study.

PCA, according to Pallant (2013), is a data reduction

technique that looks for a way a huge collection of data may

be reduced or summarized. The purpose of this study,

however, was to single out variables that did not correlate

with each other even though they belong to the same

category theoretically and explain why this might have been

the case.

PCA produces a series of tables and numbers that enable

researchers to find answers to their questions. Two of the

tables include measures of sampling adequacy and measures

of PCA appropriateness. An important number is determinant

of the correlation matrix. This number checks for the

existence of multicolinarity (correlations above .8 between

variables) and singularity (perfect correlations between

variables). SPSS provides this value at the bottom of the

correlation matrix. Determinant’s value should be

significantly different from zero.

PCA also produces two very informative plots. The first of

these plots is called the Scree plot and the other is called the

Component or Factor plot. The Scree plot tells us which

variable or groups of variables are statistically important and

should be retained. The component plot, however, represents

loadings of the variables on the components after they are

extracted.

4. Data Analysis

The questionnaire used in this study was a five-point

Likert scale questionnaire with 30 questions. The reliability

of this data collection instrument was r=.803, as shown in

Table 4.1 below. Pallant (2013) suggests that r values

above .70 are large enough for the reliability of

questionnaires.

Table 4.1. Reliability of the Questionnaire.

Reliability Statistics

Cronbach's Alpha N of Items

.803 30

PCA was used four times in this study. In the first case, it

was used to plot all of the variables on the component plot

and around the vectors. This was done to show that there

might be variables in each category that do not correlate with

each other and do not load on their respective components

rather load on the other component or fall somewhere in

between. In the second case, PCA was run without the

confounding variables, i.e., non-correlating variables. The

confounding variables were the ones whose strength of

correlation with any other variable and the principle

component to which they belonged did not exceed .3, a

criterion set by statisticians. The second instance of running

PCA could have given us a much clearer picture of what was

going on with respect to the loadings of variables on the two

principle components. It should be kept in mind, however,

that many of the non-correlating variables, as their mean

values represented, were very important ones to which

respondents had assigned some of the greatest values. This

means that, the results needed to be interpreted with respect

to the roles that these variables play in relation to foreign

language learning success not just by whether they correlate

with other variables or not. The third and fourth occasions of

running PCAs were related to exploring the variables’

interactions within the external and internal groups. Table 4.2

represents the KMO and Bartlett’s values for the total data in

this study.

Table 4.2. KMO Test of Sampling Adequacy and Bartlett’s Sphericity.

KMO and Bartlett's Test

Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .626

Bartlett's Test of Sphericity

Approx. Chi-Square 998.899

df 435

Sig. .000

According to Field (2009), the value of KMO test of

adequacy of sampling should be above the bare minimum

of .5 for us to be able to run PCA. Bartlett’s measure, tests

whether the correlation matrix is an identity matrix (that is, if

International Journal of Language and Linguistics 2015; 3(5): 313-322 317

there is any variable that does not correlate with any other

variable) or not. Identity matrices are not appropriate for

PCA tests. So, we need some correlations between variables

for PCA to work but this correlation should not be very high

which would result in multicolinarity or singularity.

Another index to be checked for is Determinant of the

correlation. The value of Determinant of the correlation is

important for rejecting multicollinarity and singularity. This

value which is given at the bottom of the Correlation Matrix

(which is not given here for its big size) must be smaller

than .05 for us to be able to reject the existence of

multicollinarity and singularity. In the case of our data the

Determinant’s value was equal to zero.

Anti-image Covariance is another necessary index which

shows the KMO values for individual questions, that is, the

adequacy of the number of responses given to each question

in the questionnaire. These values fall on the diagonal of

Anti-image table and necessarily must be above .5. The

heavily shaded values in the portion of the Anti-image

covariance table that is given below indicate the magnitudes

of this index for individual questions in our data.

Table 4.3. Anti-image Covarince.

Anti-image

Covariance

Motivation .651 -.198 .092 -.017 -.081 .075

parental

influence -.198 .556 -.103 -.079 .057 -.104

Intelligence .092 -.103 .627 -.184 .016 .190

Teachers -.017 -.079 -.184 .632 -.075 -.085

Attitude -.081 .057 .016 -.075 .556 -.096

social class .075 -.104 .190 -.085 -.096 .618

The Total Variance Explained table is an additional table

that shows the eigenvalues of the variables. Eigenvalue can

be conceived of as the ratio of the length of the data in a

scatter plot to its breadth represented by perpendicularly

crisscrossing lines. The larger this value is the more loading

it can be concluded to have on one of the principle

components. According to Field (2009), the table of Total

Variance Explained lists the eigenvalues associated with each

linear variable, i.e., each question in the questionnaire, before

and after extraction and also after rotation. The eigenvalue

associated with each variable represents the amount of

variance explained by that variable.

Variables in the TVE table are listed in a descending order

with variables on top having the largest eigenvalues. In the

Extraction Sums of Squared Loadings column, the only

difference with the previous column is that the values for the

discarded variables are ignored. The final column displays

the eigenvalues of the variables after they are rotated. Direct

Oblimin rotation was the procedure used in this analysis

because there were significant correlations between some

variables belonging to different components. Verimax is the

best choice when variables loadings on different components

do not correlate with each other significantly. The following

table shows the Rotation Sums of Squared Loadings after

rotation was done. Please note that like the Anti-image

Covariance table, only the top part of the table is presented

here. The deleted section only shows the initial eigenvalues

of the remaining variables.

Table 4.4. Eigenvalues of Variables after They Are Rotated.

Component Initial Eigenvalues Extraction Sums of Squared Loadings Rotation Sums of Squared Loadings

Total % of Variance Cumulative % Total % of Variance Cumulative % Total % of Variance Cumulative %

1 4.665 15.551 15.551 4.665 15.551 15.551 3.403 11.343 11.343

2 2.133 7.109 22.659 2.133 7.109 22.659 3.395 11.316 22.659

3 2.033 6.775 29.435

4 1.838 6.126 35.561

5 1.598 5.328 40.889

6 1.469 4.898 45.787

Field (2009) argues that PCA “works on the initial

assumption that all variance is common; therefore, before

extraction the communalities are all 1” (p. 661). However,

once principle components are extracted, we have a better

idea of how much variance is common. For example, in the

part of the Communalities table that follows, we can say

that .114% of the variance associated with the first question

is common or shared variance.

Table 4.5. Communalities of Variables.

Communalities

Initial Extraction

Motivation 1.000 .114

parental influence 1.000 .331

Intelligence 1.000 .141

Teachers 1.000 .233

Attitude 1.000 .142

As it was said, SPSS provides a couple of graphs in

addition to tables in PCA. One of the very useful graphs

produced is the Scree plot. This plot shows the importance of

each variable schematically. Usually, there are one or a few

variables that have substantial loadings on the principle

components and occupy the highest points on the Scree plot.

Variables with relatively low eigenvalues, of which usually

there are many, fall after the sharp descent or point of

inflexion and tail off with a mild slope. The Scree plot that

follows shows the loadings of variables in this study

diagrammatically. There are two points of inflexion, as it can

be seen, one after the second variable and the other after the

third variable. This suggests that there might have been two

principle components, as we had speculated initially. The

eigenvalue of the variable with the highest loading is close to

5 while the eigenvalues of variables in the first and second

inflexion points are around 2. These variables can be

identified by the means of the responses to questions if they

are arranged from the biggest to the smallest. In the case of

318 Soheil Mahmoudi and Asgar Mahmoudi: Internal and External Factors Affecting

Learning English as a Foreign Language

our data, the first two largest loadings belonged to motivation

and teachers.

Figure 4.1. Loadings of the variables.

Another important graph that is very informative is the

Component Plot, alternatively called Factor Plot. This plot is

especially easy to draw when there are only two principle

components, because the loadings of variables on them can

be represented by two vertical and horizontal axes or vectors.

In a two-dimensional component plot, variables that relate to

each principle component are plotted around those

components represented by the axes. The coordinates of each

variable represent the strength of the relationship between

that variable and each of the components. That is, the

position of each variable says how much it correlates with

each of the components. The axes lines range from -1 to 1

which are the outer limits of a correlation coefficient (Field,

2009). In Figure 4.2 below a component plot for the data in

this study is given. As it is evident, a cloud of dots covers the

area between the two axes on the right top most quarter of the

graph. The dots falling on the middle of the cloud represent

the variables that not only do not correlate with a particular

principle component strongly enough but also do not

correlate with other variables of their own type. This finding

is confirmed by looking at the Correlations Matrix and

provides evidence for accepting our both first and second

hypotheses. Remember that our first hypothesis stated that

not all variables that are theoretically considered to be

internal or external necessarily correlate significantly with

variables of their own type. In our Correlation Matrix (which

is not given here for its large size) there were at least four

internal variables and seven external variables that did not

correlate with variables of their own type significantly. This

is realizable from the fact that on the component plot

variables are not clustered tightly around their respective

components. And our second hypothesis stated that there are

no internal and external variables that correlate significantly

with each other or with the principle component to which

they do not belong. Again, considering the odd and even

numbers around each vector that represent internal and

external variables, respectively, we can see that there are

variables of each type clustering around both of the axes.

This means that there, indeed, had been variables of different

types that correlated significantly with each other and with

the principle component to which they did not belong.

Figure 4.2. Components and their contributing variables.

Strangely enough, the non-correlating variables are some

of the most important variables as rated by the respondents.

The results obtained in the Rotated Component Matrix below

too show variables of different nature having loadings on the

Component to which they do not belong conceptually.

Table 4.6. Rotated Loadings of Variables on Components.

Rotated Component Matrixa

Component

1 2

Beliefs .738

job market .670

Politics .621

risk-taking ability .506

Personality .499

social class .419

Press .417

first language .338 .310

Autonomy .337

analytical perception .334

Friends .334

degree of hopefulness

group work .610

early start .575

Films .524

parental influence .495

Anxiety .494

participation .352 .443

Age .438

Teachers .431

competitiveness .413

openness to innovation and new

methods .385

Intelligence .373

L2 curriculum .335

Motivation .331

Institution .313

Internet

Attitude

teaching resources

persistence

The non-correlating variables are questions addressing

motivation, intelligence, teachers, social class, autonomy,

institutions, friends, persistence, the Internet, teaching

resources, and L2 curriculum with mean values of 4.80, 3.97,

4.18, 3.42, 3.52, 3.5, 3.15, 4.11, 3.80, 3.68, 3.17, respectively.

Some of these variables are internal and others external. The

International Journal of Language and Linguistics 2015; 3(5): 313-322 319

empty spaces are there because correlations between the

variable and the principle components had been below. 3.

This state of affairs, however, does not reject the idea that,

generally speaking, there are two principle components. To

show that there are specific groups of variables that load on

just one component we can discard our non-correlating

variables and run PCA one more time. The Component Plot

and the Rotated Component Matrix resulting from this

pruning process will be more revealing as can be seen below

in Figure 4.3 and Table 4.7.

Figure 4.3. Components and their contributing variables after pruning.

Table 4.7. Loadings of Variables on Components after Pruning.

Rotated Component Matrixa

Component

1 2

group work .698

Films .672

early start .594

parental influence .555

participation .529

Age .525

first language .478

Press .389

Beliefs .720

job market .647

risk-taking ability .608

personality .569

Politics .527

degree of hopefulness .419

Attitude .400

To test our third research hypothesis we needed to run an

Independent-samples t-test to find out if different variable

types had been treated, i.e., rated, differently by EFL learners

with respect to their importance or not. Before running this

test it would be better for us to know that the grand mean of

the scores assigned to the internal variables was 3.767 and

the grand mean of the scores assigned to the external

variables was 3.499 and this is while only three external

variables, namely, politics, friends, and parental influence, by

mean scores of 2.71, 3.03, and 3.15, respectively, gathered

low mean scores compared to other external variables. If

students had enough information about the role of politicians

in setting educational goals, the role that parents play in their

children’s learning of English as a foreign language, and in

developing positive attitudes toward this language, and the

role that peers play in many cooperative learning contexts,

their responses could have favored external variables. Table

4.8 shows the lack of significant difference in the importance

assigned to internal and external principle components by

respondents in this study.

Table 4.8. Independent-samples T-test Comparing Means of Internal and

External Variables.

Independent Samples Test

Levene's Test for

Equality of Variances

t-test for Equality of

Means

F Sig. t df Sig. (2-tailed)

Equal variances

assumed .990 .328 1.897 28 .068

The result obtained from running Independent-samples t-

test convinces us that the mean values of the students’

responses to different types of questions did not differ from

each other significantly which pushes us to accept our third

hypothesis, that is, the importance of internal and external

variables from the respondents’ perspectives had almost been

the same.

Another thing that could be investigated in relation to the

questions in the questionnaire was if the internal and external

components themselves are divisible. Answering this

question could enable us to either reject our fourth research

hypothesis which was: Internal and external variables are

divisible to cognitive and affective and curricular and

environmental, respectively, or accept it. Since, we have

already become familiar with the workings of PCA, it would

not be very difficult to interpret the results reported below

with minimum explanation needed.

To begin with, the internal and external variables needed to

be separated from each other. It was said in section three that

the questionnaire was divisible to odd-even variables with

odd numbers representing the internal and even numbers

representing the external variables.

Running a PCA on the internal variables returned a non-

significant Determinant value of .124 below the Correlation

Matrix. This meant that the assumptions of multicolinarity

and singularity were not violated and that the data was

appropriate for running PCA. The KMO test, also, with a

value of .641 was above the bare minimum of .5, which is the

lowest limit for the adequacy of sampling. The significance

level of the Bartlet’s Test of Sphericity, likewise, was

below .05 which meant that all variables correlated with each

other to some extent. Information related to the KMO and

Bartlett’s Sphericity tests are given in Table 4.9 below.

Communalities table for internal variables is ignored here,

as it will be ignored for the external variables, because the

kind of information they provide for us is not vital to our

understanding of the remaining of this study. Table 4.10

represents the eigenvalues or loadings of the variables on the

320 Soheil Mahmoudi and Asgar Mahmoudi: Internal and External Factors Affecting

Learning English as a Foreign Language

principle components before and after extraction and after

rotation. There are many variables that have eigenvalues

above 1 in this table, but since we had set PCA to extract

only two variables for us after rotation, only the loadings of

two variables are given in the third column.

Table 4.9. Tests of Adequacy of Sampling and Sphericity for Internal

Variables.

KMO and Bartlett's Test

Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .641

Bartlett's Test of Sphericity

Approx. Chi-Square 267.404

df 105

Sig. .000

It was said that KMO is a measure of sampling adequacy,

but this measure is used at two stages: first, to see if the

number of respondents to the questionnaire was enough and

second, to see if every question was attempted by enough

respondents. While the KMO and Bartlett’s Test of Sphericity

provide answer for the first question, the answer to the

second question can be found in the Anti-image Covariances

table. The values of the Anti-image Covariances should also

go beyond .5 for PCA to work. These values for internal

variables are represented by the heavily shaded areas on the

diagonal of the Anti-image Covariances table below.

Table 4.10. Eigenvalues of Internal Variables after They Are Rotated.

Total Variance Explained

Component Initial Eigenvalues Extraction Sums of Squared Loadings Rotation Sums of Squared Loadingsa

Total % of Variance Cumulative % Total % of Variance Cumulative % Total

1 2.857 19.045 19.045 2.857 19.045 19.045 2.695

2 1.446 9.639 28.684 1.446 9.639 28.684 1.848

3 1.378 9.185 37.869

4

5

1.273

1.147

8.488

7.644

46.357

54.001

Table 4.11. Anti-image Covariances.

Q1 Q3 Q5 Q7 Q9 Q11 Q13 Q15 Q17 Q19 Q21 Q23 Q25 Q27 Q29

Anti-image

Covariances

Q1 .551a .057 -.124 -.006 -.113 .179 -.109 -.190 .017 -.110 -.040 .111 -.094 .060 -.092

Q3 .057 .604a -.056 -.088 -.101 -.044 -.070 .107 -.001 -.021 -.081 .138 -.077 .059 -.078

Q5 -.124 -.056 .658a -.055 -.234 -.202 .056 .002 .093 -.144 .032 -.095 -.304 .088 -.114

Q7 -.006 -.088 -.055 .648a -.157 .080 .026 -.061 -.068 .103 -.220 -.048 -.101 -.088 -.236

Q9 -.113 -.101 -.234 -.157 .630a -.119 .106 -.101 .121 -.082 -.005 -.011 .138 -.282 .079

Q11 .179 -.044 -.202 .080 -.119 .699a -.147 -.151 -.048 -.140 -.114 .163 -.097 -.200 -.023

Q13 -.109 -.070 .056 .026 .106 -.147 .666a .036 -.011 -.143 -.100 -.182 -.097 .021 -.065

Q15 -.190 .107 .002 -.061 -.101 -.151 .036 .683a -.162 .067 -.148 .054 -.047 .034 .039

Q17 .017 -.001 .093 -.068 .121 -.048 -.011 -.162 .664a -.352 -.112 -.107 -.076 -.221 -.134

Q19 -.110 -.021 -.144 .103 -.082 -.140 -.143 .067 -.352 .676a -.154 -.068 .071 .060 .071

Q21 -.040 -.081 .032 -.220 -.005 -.114 -.100 -.148 -.112 -.154 .694a .001 .171 .024 .075

Q23 .111 .138 -.095 -.048 -.011 .163 -.182 .054 -.107 -.068 .001 .541a -.115 -.058 -.093

Q25 -.094 -.077 -.304 -.101 .138 -.097 -.097 -.047 -.076 .071 .171 -.115 .580a -.123 .092

Q27 .060 .059 .088 -.088 -.282 -.200 .021 .034 -.221 .060 .024 -.058 -.123 .619a .118

Q29 -.092 -.078 -.114 -.236 .079 -.023 -.065 .039 -.134 .071 .075 -.093 .092 .118 .495a

Measures of Sampling Adequacy(MSA)a

Table 4.12. Pattern of Internal Variables Showing their Loadings on

Principle Components.

Pattern Matrixa

Component

1 2

Q9 .724 -.368

Q11 .660

Q27 .567

Q5 .550

Q15 .483

Q7 .385

Q21 .371

Q25 .326

Q3

Q1

Q13 .629

Q23 .585

Q17 .314 .501

Q29 .466

Q19 .361 .431

Up to this point, we have been talking about the conditions

that should be met before running any PCAs. However, to

understand about the components and the loadings of

variables on them we need to look at Pattern and Rotated

Component matrices and Scree and component plots. Table

4.12 shows the Pattern Matrix for the internal variables.

The scree plot shows information on the loadings of the

internal variables. The most important graph for us, however,

is the component plot that follows it. The component plot

shows whether variables cluster around one component or the

other or are distributed unpredictably between the two

components.

If we call the two principle components of internal

variables as affective and cognitive, we can see that only

three of the variables have clustered around the vertical

component and the rest are bundled around the horizontal

axis. This means that even with more specific categorization

of variables the respondents’ answers to the questions do not

International Journal of Language and Linguistics 2015; 3(5): 313-322 321

converge or diverge on the basis of variables’ nature.

Ignoring other details, the Component plot for the external

variables shows even a more confusing picture of the

loadings of this type of variables on the curricular and

environmental components, as can be seen in Figure 4.6.

Figure 4.4. Scree plot for the internal variables.

Figure 4.5. Component plot for the internal variables.

Figure 4.6. Component plot for the external variables.

These findings drive us to the conclusion that our fourth

hypothesis concerning the divisibility of the internal and

external variables to cognitive and affective and curricular

and environmental, respectively, is not tenable with respect to

the amount of importance assigned to them although this

division might be valid theoretically. That is, students do not

accord much importance to a variable simply because it is

internal or external, rather they look back at their own

language experience and try to figure out what it was that

made them a successful learner or what discouraged them

from learning be it internal or external.

5. Conclusion and Discussion

The finding that there were two principle components is not

very important because we already knew about it. What is

important is that to determine the elements of success in foreign

language learning one should not exclusively concentrate on

variables that load on one principle component or the other but

to concentrate on both of them and even variables that load on

none of them heavily. Alternatively, one could focus on variables

that although belonging to a particular category, do not correlate

highly with variables of the same nature.

One reason for this is that variables that theoretically are

related to each other and fall in the same category do not

necessarily correlate highly with each other and even with the

principle component that represents them. For example, in the

case of this study motivation, which had the highest mean

score among all of the variables, did not correlate highly with

the internal variables like intelligence and persistence. It did

not load heavily on the internal principle component either, as

can be seen in Table 4.6. On the other hand, ‘teachers’ variable

which was the second most important variable from the

respondents’ perspective did not correlate highly with teaching

resources and institutions and did not load heavily on the

external component either, as is visible again in Table 4.6.

These findings altogether suggest that categories should not be

the basis of our judgment by saying that, for example, internal

variables are more important than external variables merely

because these are emphasized more in the literature. What

matters, is paying balanced attention to both categories of

variables and exploring their effects.

References

[1] Attitude. (2009). In Encyclopedia Wikipedia online. Retrieved from http://en.wikipedia,org/second_language_acquisition/attitue.

[2] Brown, H. D. (2007). Principles of Language Learning and Teaching (5th ed.). New York: Pearson Education, Inc.

[3] Brown, J. D. (1995). The elements of language curriculum. Boston, Massachusetts: Heinle & Heinle Publishers.

[4] Cheng, H., & Dӧrnyei, Z. (2007). The use of motivational strategies in language instruction: the case of EFL teaching in Taiwan. Innovation in Language Learning and Teaching, 1(1), 153–174.

322 Soheil Mahmoudi and Asgar Mahmoudi: Internal and External Factors Affecting

Learning English as a Foreign Language

[5] Cohen, A. D. (2010). Focus on language learner: Style, strategies, and motivation. In N. Schmitt (Ed.), An introduction to applied linguistics (2nd ed.). London: Hodder & Stoughton Ltd.

[6] Dӧrnyei, Z. and Csizer, K. (1998). Ten commandments for motivating language learners: Results of an empirical study. Language Teaching Research, 2(3), 203–229.

[7] Ellis, R. (2008). The study of second language acquisition (2nd ed.). New York: Oxford University Press.

[8] Field, A. (2009). Discovering statistics using SPSS (3rd ed.). London: Sage Publications Ltd.

[9] Gardner, R. C. (1985). Social psychology and second language learning: The role of attitudes and motivation. London: Edward Arnold.

[10] Gass, S., & Selinker, L. (2008). Second language acquisition (3rd ed.). New York: Routledge.

[11] Ismaili, M., (2013). The effectiveness of using movies in the EFL classroom: A study conducted at South East European University. Academic Journal of Interdisciplinary Studies, 2 (4), 110–118.

[12] Karim, K. (2003). First language (L1) influence on second language (L2) reading: The role of transfer. Proceedings of the 19th Northwest Linguistics Conference. (pp. 324–347). Victoria, B.C.: Victoria.

[13] Khanchali, M., & Zidat, A. (2011). The impact of the Internet on the development of students’ writing. Revue des Sciences Humaines 21, 51–62.

[14] Kinginger, C., & Farrell, K. (2004). Assessing development of meta-pragmatic awareness in study abroad. Frontiers: The Interdisciplinary Journal of Study Abroad, 10, 19–42.

[15] Lewis, M., & Hill, J. (1985). Practical techniques for language teaching. London: Language Teaching Publications.

[16] Madrid, D. (1995). Interanl and external factors in language teaching. Actas de las 1I Jornadas de Estudios Ingleses. Universidad de Jaen, 2 (2), 59–82.

[17] Nation, I. S. P., & Macalister, J (2010). Language curriculum design. New York: Routledge.

[18] Nunan, D. (1988). Syllabus design. New York: Oxford University Press.

[19] Nurul Islam, M. (2011). Independent English learning through the Internet. Journal of Language Teaching and Research, 2 (5), 1080–1085.

[20] Ortega, L. (2009). Understanding second language acquisition. London: Hodder Education.

[21] Pallant, J. (2013). SPSS survival manual. New York: Open University Press.

[22] Richards, J. C. (2001). Curriculum development in language teaching. New York: Cambridge University Press.

[23] Robinson, P., & Ellis, N. C. (2008). An introduction to cognitive linguistics, second language acquisition, and language instruction. In P. Robinson, & N. C. Ellis (Eds.). Handbook of cognitive linguistics and second language acquisition. New York: Routledge.

[24] Segalowitz, N. (1986). Skilled reading in the second language. In J. Vaid (Ed.). Language Processing in bilinguals: Psycholinguistic and Neuropsychological Perspectives. Hillsdale, NJ: Lawrence Erlbaum.

[25] Siegel, J. (2003). Social context. In C. J. Doughthy, & M. H. Long (Eds.). The handbook of second language acquisition. Oxford: Blackwell Publishing Ltd.

[26] Stipek, D. J. (2002). Motivation to learn: Integrating theory and practice (4th ed.). Boston, MA: Allyn and Bacon.

[27] White, R. (1988). The ELT curriculum: Design, innovation and management. Oxford: Basil Blackwell Ltd.

[28] Woolfolk, A. E., Winne, P. H., & Perry, N. E. (2003). Educational Psychology (2nd ed.). Toronto: Pearson Education Canada Inc.